Lachancea sp. yHQL494, whole genome shotgun sequencing project
Lachancea sp. yHQL494, whole genome shotgun sequencing project
accharomycetes↗
SEARCH · Search NASA
Results for “shotgun sequencing”
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Aureobasidium pullulans strain CRUB 1823, whole genome shotgun sequencing project
Aureobasidium pullulans strain CRUB 1823, whole genome shotgun sequencing project.
Ascomycota/genetics↗
Sea cucumber ( Holothuria glaberrima ) intestinal microbiome dataset from Puerto Rico, generated by shotgun sequencing
The sea cucumber (H. glaberrima) is a species found in the shallow waters near coral reefs and seagrass beds in Puerto Rico. To characterize the microbial taxonomic composition and functional profiles present in the sea cucumber, total DNA was obtained from their intestinal system, fosmid libraries constructed, and subsequent sequencing was performed. The diversity profile displayed that the most predominant domain was Bacteria (76.56 %), followed by Viruses (23.24 %) and Archaea (0.04 %). Within the 11 phyla identified, the most abundant was Proteobacteria (73.16 %), followed by Terrabacteria group (3.20 %) and Fibrobacterota, Chlorobiota, Bacteroidota (FCB) superphylum (1.02 %). The most abundant species were Porvidencia rettgeri (21.77 %), Pseudomonas stutzeri (14.78 %), and Alcaligenes faecalis (5.00 %). The functional profile revealed that the most abundant functions are related to transporters, MISC (miscellaneous information systems), organic nitrogen, energy, and carbon utilization. The data collected in this project on the diversity and functional profiles of the intestinal system of the H. glaberrima provided a detailed view of its microbial ecology. These findings may motivate comparative studies aimed at understanding the role of the microbiome in intestinal regeneration.
59 BASIC BIOLOGICAL SCIENCES↗
Saccharomycopsis sp. KB-2023a isolate UFMG-CM-Y6991, whole genome shotgun sequencing project
Three yeast isolates were obtained from soil and rotting wood samples collected in an Amazonian rainforest biome in Brazil. Comparison of the intergenic spacer 5.8S region and the D1/D2 domains of the large subunit rRNA gene showed that the isolates represent a novel species of the genus Saccharomycopsis. A tree inferred from the D1/D2 sequences placed the novel species near a subclade containing Saccharomycopsis lassenensis, Saccharomycopsis fermentans, Saccharomycopsis javanensis, Saccharomycopsis babjevae, Saccharomycopsis schoenii and Saccharomycopsis oosterbeekiorum, but with low bootstrap support. In terms of sequence divergence, the novel species had the highest identity in the D1/D2 domains with Saccharomycopsis capsularis, from which it differed by 36 substitutions. In contrast, a phylogenomic analysis based on 1061 single-copy orthologs for a smaller set of Saccharomycopsis species whose whole genome sequences are available indicated that the novel species represented by strain UFMG-CM-Y6991 is phylogenetically closer to Saccharomycopsis fodiens and Saccharomycopsis sp. TF2021a (=Saccharomycopsis phalluae). The novel yeast is homothallic and produces asci with one spheroidal ascospore with an equatorial or subequatorial ledge. The name Saccharomycopsis praedatoria sp. nov. is proposed to accommodate the novel species. The holotype of Saccharomycopsis praedatoria is CBS 16589(T). The MycoBank number is MB849369. S. praedatoria was able to kill cells of Saccharomyces cerevisiae by means of penetration with infection pegs, a trait common to most species of Saccharomycopsis.
Amazonian Forest↗
Pichia sp. yHMH446, whole genome shotgun sequencing project
A novel budding yeast species was isolated from a soil sample collected in the United States of America. Phylogenetic analyses of multiple loci and phylogenomic analyses conclusively placed the species within the genus Pichia. Strain yHMH446 falls within a clade that includes Pichia norvegensis, Pichia pseudocactophila, Candida inconspicua, and Pichia cactophila. Whole genome sequence data were analyzed for the presence of genes known to be important for carbon and nitrogen metabolism, and the phenotypic data from the novel species were compared to all Pichia species with publicly available genomes. Across the genus, including the novel species candidate, we found that the inability to use many carbon and nitrogen sources correlated with the absence of metabolic genes. Based on these results, Pichia galeolata sp. nov. is proposed to accommodate yHMH446(T) (=NRRL Y-64187 = CBS 16864). This study shows how integrated taxogenomic analysis can add mechanistic insight to species descriptions.
genome sequence↗
Tilletia maclaganii strain Tm001 NY21, whole genome shotgun sequencing project
GenBank: JBQVXD000000000.1
Basidiomycota↗
Genome Extraction from Shotgun Metagenome Sequence Data
Uncultivated Bacteria and Archaea comprise the vast majority of species on Earth, but obtaining their genomes directly from the environment, using shotgun sequencing, has only recently become possible. To realize the hope of capturing Earth’s microbial genetic complement, technologies that accelerate recovery of high-quality genomes are necessary. We present a series of analysis steps and data products for the extraction of high quality metagenome-assembled genomes (MAGs) from microbiomes using the U.S. Department of Energy Systems Biology Knowledgebase (KBase) platform (http://www.kbase.us/). In KBase, the process is end-to-end, allowing a user to go from the initial sequencing reads all the way through to MAG genomes, which can then be analyzed with other KBase capabilities such as phylogenetic placement, functional assignment, metabolic modeling, pangenome functional profiling, RNA-Seq, and others. While portions of such capabilities are individually available from other resources, the combination of the intuitive usability, data interoperability, and integration of tools in a freely available compute resource makes KBase a uniquely powerful platform for obtaining MAGs from microbiomes. While this workflow offers tools for each of the key steps in the genome extraction process, it also provides a scaffold that can be easily extended, with additional MAG recovery and analysis tools, via the KBase SDK (Software Development Kit).
Chivian, Dylan↗
Hybridization capture sequencing for Vibrio spp. and associated virulence factors
ABSTRACT Proliferation ofVibriospp. in aquatic ecosystems is associated with climate change and, concomitantly, increased incidence of vibriosis. They are autochthonous to aquatic environments globally, but traditional metagenomic methods for detecting and typing pathogenicVibriospp. are challenged by their presence in relatively low abundance and ability to persist in a viable but nonculturable state. In the study reported here, hybridization capture sequencing (HCS) was employed to profile low-abundanceVibriospp. in environmental samples. The HCS panel targeted a family of molecular chaperones (CPN60) specific to 69Vibriospp. and 162Vibrio-specific virulence factors. This approach was evaluated in parallel with traditional whole-community shotgun sequencing in a metagenomic analysis of water and oyster samples collected from the Chesapeake Bay. In addition,Vibrio parahaemolyticusandVibrio vulnificusstrains isolated from the samples were subjected to whole-genome sequencing to determine the genetic characteristics of pathogenicVibriospp. circulating in an aquatic environment. HCS, employed to determine the incidence and characterization of specificVibriospp., yielded significantly greater metagenomic insight, notably a variety of otherVibriospp., including detection ofVibrio cholerae,Vibrio fluvialis, andVibrio aestuarianus, in addition toVibrio parahaemolyticusandVibrio vulnificus, and also important virulence factors not detectable using traditional molecular methods. Thus, pathogenicVibriospp. in aquatic ecosystems may be far more common than currently understood. It is concluded that environmental surveillance should include HCS, a valuable tool for the detection and characterization of pathogenic agents in aquatic ecosystems, notably vibrios. IMPORTANCE The increasing prevalence of pathogenicVibriospp. in aquatic ecosystems, driven by climate change, is closely linked to a rise in cholera and vibriosis cases, emphasizing the need for improved environmental surveillance. Vibrios are naturally occurring in aquatic environments globally, but traditional metagenomic methods for detecting and typing pathogenicVibriospp. are challenged by their presence in relatively low abundance and ability to persist in a viable but nonculturable state. In the study reported here, hybridization capture sequencing was employed to profile low-abundanceVibriospp. in metagenomic samples, namely water and oysters collected from the Chesapeake Bay. This approach was evaluated in parallel with traditional whole-community shotgun sequencing and whole-genome sequencing ofVibrio parahaemolyticusandVibrio vulnificusstrains isolated from the samples. Results suggest pathogenicVibriospp. in aquatic ecosystems may be far more common than currently understood, when multiple methods are considered for environmental surveillance.
Microbiology↗
Determining the Reliability of Measuring Mercury Cycling Gene Abundance with Correlations with Mercury and Methylmercury Concentrations
Methylmercury (MeHg) is a bioaccumulative toxic contaminant in many ecosystems but factors governing its production are poorly understood. Recent work has shown that the anaerobic microbial conversion of mercury (Hg) to MeHg requires the Hg-methylation genes hgcAB and that these genes can be used as biomarkers in PCR-based estimators of Hg-methylator abundance. In an effort to determine reliable methods for assessing hgcA abundance and diversity and linking it to MeHg concentrations, multiple approaches were compared including metagenomic shotgun sequencing, 16S rRNA gene pyrosequencing and cloning/sequencing hgcAB gene products. Hg-methylator abundance was also determined by quantitative hgcA qPCR amplification and metaproteomics for comparison to the above measurements. Samples from eight Hg-contaminated sites were examined over a range of total Hg (HgT; 0.03–14 mg kg-1 wet wt. soil) and MeHg (0.05–27 µg kg-1 wet wt. soil) concentrations. In the metagenome and amplicon sequencing of hgcAB diversity, the Deltaproteobacteria were the dominant Hg-methylators, while Firmicutes and methanogenic Archaea were typically ~50% less abundant. This was consistent with metaproteomics estimates where the Deltaproteobacteria were consistently higher. The 16S rRNA gene pyrosequencing did not have sufficient resolution to identify hgcAB+ species. Metagenomic and hgcAB results were similar for Hg-methylator diversity and showed a tractable relationship between Hg-methylating bacteria and soil Hg concentrations among similar environmental samples. However, clade-specific qPCR-based approaches for hgcA are only appropriate when comparing the abundance of a particular clade across various samples.
59 BASIC BIOLOGICAL SCIENCES↗
synDNA—a Synthetic DNA Spike-in Method for Absolute Quantification of Shotgun Metagenomic Sequencing
The synDNAs designed in this study enable accurate and reproducible measurements of absolute amount and fold changes of bacterial species in complex microbial communities. The method proposed here is versatile and promising as it can be applied to bacterial communities or genomic features like genes and operons, in addition to being easily adaptable by other research groups at a low cost.
59 BASIC BIOLOGICAL SCIENCES↗
A pipeline for targeted metagenomics of environmental bacteria
Background:Metagenomics and single cell genomics provide a window into the genetic repertoire of yet uncultivated microorganisms, but both methods are usually taxonomically untargeted. The combination of fluorescence in situ hybridization (FISH) and fluorescence activated cell sorting (FACS) has the potential to enrich taxonomically well-defined clades for genomic analyses. Methods:Cells hybridized with a taxon-specific FISH probe are enriched based on their fluorescence signal via flow cytometric cell sorting. A recently developed FISH procedure, the hybridization chain reaction (HCR)-FISH, provides the high signal intensities required for flow cytometric sorting while maintaining the integrity of the cellular DNA for subsequent genome sequencing. Sorted cells are subjected to shotgun sequencing, resulting in targeted metagenomes of low diversity. Results: Pure cultures of different taxonomic groups were used to (1) adapt and optimize the HCR-FISH protocol and (2) assess the effects of various cell fixation methods on both the signal intensity for cell sorting and the quality of subsequent genome amplification and sequencing. Best results were obtained for ethanol-fixed cells in terms of both HCR-FISH signal intensity and genome assembly quality. Our newly developed pipeline was successfully applied to a marine plankton sample from the North Sea yielding good quality metagenome assembled genomes from a yet uncultivated flavobacterial clade. Conclusions: With the developed pipeline, targeted metagenomes at various taxonomic levels can be efficiently retrieved from environmental samples. The resulting metagenome assembled genomes allow for the description of yet uncharacterized microbial clades.
59 BASIC BIOLOGICAL SCIENCES↗
Function-driven single-cell genomics uncovers cellulose-degrading bacteria from the rare biosphere
Assigning a functional role to a microorganism has historically relied on cultivation of isolates or detection of environmental genome-based biomarkers using a posteriori knowledge of function. However, the emerging field of function-driven single-cell genomics aims to expand this paradigm by identifying and capturing individual microbes based on their in situ functions or traits. To identify and characterize yet uncultivated microbial taxa involved in cellulose degradation, we developed and benchmarked a function-driven single-cell screen, which we applied to a microbial community inhabiting the Great Boiling Spring (GBS) Geothermal Field, northwest Nevada. Our approach involved recruiting microbes to fluorescently labeled cellulose particles, and then isolating single microbe-bound particles via fluorescence-activated cell sorting. The microbial community profiles prior to sorting were determined via bulk sample 16S rRNA gene amplicon sequencing. The flow-sorted cellulose-bound microbes were subjected to whole genome amplification and shotgun sequencing, followed by phylogenetic placement. Next, putative cellulase genes were identified, expressed and tested for activity against derivatives of cellulose and xylose. Alongside typical cellulose degraders, including members of the Actinobacteria, Bacteroidetes, and Chloroflexi, we found divergent cellulases encoded in the genome of a recently described candidate phylum from the rare biosphere, Goldbacteria, and validated their cellulase activity. As this genome represents a species-level organism with novel and phylogenetically distinct cellulolytic activity, we propose the name Candidatus ‘Cellulosimonas argentiregionis’. We expect that this function-driven single-cell approach can be extended to a broad range of substrates, linking microbial taxonomy directly to in situ function.
59 BASIC BIOLOGICAL SCIENCES↗
Review, Evaluation, and Directions for Gene-Targeted Assembly for Ecological Analyses of Metagenomes
Shotgun metagenomics has greatly advanced our understanding of microbial communities over the last decade. Metagenomic analyses often include assembly and genome binning, computationally daunting tasks especially for big data from complex environments such as soil and sediments. In many studies, however, only a subset of genes and pathways involved in specific functions are of interest; thus, it is not necessary to attempt global assembly. In addition, methods that target genes can be computationally more efficient and produce more accurate assembly by leveraging rich databases, especially for those genes that are of broad interest such as those involved in biogeochemical cycles, biodegradation, and antibiotic resistance or used as phylogenetic markers. Here, we review six gene-targeted assemblers with unique algorithms for extracting and/or assembling targeted genes: Xander, MegaGTA, SAT-Assembler, HMM-GRASPx, GenSeed-HMM, and MEGAN. We tested these tools using two datasets with known genomes, a synthetic community of artificial reads derived from the genomes of 17 bacteria, shotgun sequence data from a mock community with 48 bacteria and 16 archaea genomes, and a large soil shotgun metagenomic dataset. We compared assemblies of a universal single copy gene (rplB) and two N cycle genes (nifH and nirK). We measured their computational efficiency, sensitivity, specificity, and chimera rate and found Xander and MegaGTA, which both use a probabilistic graph structure to model the genes, have the best overall performance with all three datasets, although MEGAN, a reference matching assembler, had better sensitivity with synthetic and mock community members chosen from its reference collection. Also, Xander and MegaGTA are the only tools that include post-assembly scripts tuned for common molecular ecology and diversity analyses. Additionally, we provide a mathematical model for estimating the probability of assembling targeted genes in a metagenome for estimating required sequencing depth.
59 BASIC BIOLOGICAL SCIENCES↗
Draft genome of multiple resistance donor plant Sinapis alba: An insight into SSRs, annotations and phylogenetics
Sinapis alba is a wild member of the Brassicaceae family reported to possess genetic resistance against major biotic and abiotic stresses of oilseed brassicas. However, the resistance nature of S. alba was not exploited generously due to the unavailability of usable genome sequences in public databases. Therefore, the present study was conducted to assemble the first draft genome from raw whole genome shotgun sequences with annotation and develop simple sequence repeat markers for molecular genetics and marker-assisted breeding. Results The raw genome sequences had 96x coverage on the Illumina platform with 170 Gbp data. The developed assembly by SOAPdenovo2 has ~459 Mbp genome size covered in 403,423 contigs with an average size of 1138.04 bp. The assembly was BLASTX with Arabidopsis thaliana which showed 32.9% positive hits between both plants. The top hit species distribution analysis showed the highest similarity with A. thaliana. A total of 809,597 GO level annotations were recorded after BLASTX results, and 34,012 sequences were annotated with different enzyme codes grouped under seven classes. The gene prediction tool AUGUSTUS identified 113,107 probable genes with an average size of 684 bp. The biochemical pathway annotation assigned 16,119 potential genes to 152 KEGG maps and 1751 enzyme codes. The development of potential SSRs from the de-novo assembly yielded 70731 unique primer pairs. Out of 159 randomly selected SSR markers for validation, 149 successfully amplified in S. alba. However, 10 SSR markers did not amplify during the validation experiment. Conclusion The annotated genome assembly with a large number of SSRs was developed in the present study. To the best of our knowledge, this is the first report of S. alba genome assembly development, annotation, and SSRs mining to date. The data presented here will be a very important resource for future crop improvement programs, especially for resistant breeding.
59 BASIC BIOLOGICAL SCIENCES↗
Accurate and complete genomes from metagenomes
Genomes are an integral component of the biological information about an organism; thus, the more complete the genome, the more informative it is. Historically, bacterial and archaeal genomes were reconstructed from pure (monoclonal) cultures, and the first reported sequences were manually curated to completion. However, the bottleneck imposed by the requirement for isolates precluded genomic insights for the vast majority of microbial life. Shotgun sequencing of microbial communities, referred to initially as community genomics and subsequently as genome-resolved metagenomics, can circumvent this limitation by obtaining metagenome-assembled genomes (MAGs); but gaps, local assembly errors, chimeras, and contamination by fragments from other genomes limit the value of these genomes. Here, we discuss genome curation to improve and, in some cases, achieve complete (circularized, no gaps) MAGs (CMAGs). To date, few CMAGs have been generated, although notably some are from very complex systems such as soil and sediment. Through analysis of about 7000 published complete bacterial isolate genomes, we verify the value of cumulative GC skew in combination with other metrics to establish bacterial genome sequence accuracy. The analysis of cumulative GC skew identified potential misassemblies in some reference genomes of isolated bacteria and the repeat sequences that likely gave rise to them. We discuss methods that could be implemented in bioinformatic approaches for curation to ensure that metabolic and evolutionary analyses can be based on very high-quality genomes.
59 BASIC BIOLOGICAL SCIENCES↗
Transcriptomics-based Machine Learning Analysis Predicts Space-Exposed Murine Livers
Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), typically limits machine learning (ML) approaches in spaceflight studies that include radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNA-seq) data from six mouse liver GeneLab datasets (GLDS) (n ranging from 6 to 39 samples) from with a total of 81 spaceflight and ground-control samples to determine top features (i.e. genes) relevant to spaceflight including the effect of radiation exposure. RNASeq counts were normalized for each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. Redundancy and relevance were computed using the Pearson correlation and F-statistic, respectively. The top 100 MRMR features were used to predict spaceflight vs. ground-control samples using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) classifiers with 5-fold cross validation (CV). Principal component analysis (PCA) on the complete feature set versus the MRMR features shows separation between spaceflight samples and ground controls (Figure 1A). The ML-based gene sets were compared against differential gene expression results obtained with DESeq2 from individual GLDS. Using all features or randomly sampled subsets at matching set sizes with MRMR, a maximum classifier accuracy of 69% was shown on the test set over 5 folds. For all classifiers, CV training using at least the top 30 MRMR genes show minimum 89% accuracy and 0.95 AUC value on the test set over 5 folds (Figure 1B). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 295 DEGs that overlap at least two studies and 13 DEGs that overlap three studies (Figure 1C). Set analysis between the top 100 MRMR features and the DEGs showed 47 genes that overlap at least one study and 24 genes that overlap two studies. Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism which may indicate these processes in the response to spaceflight stressors. MRMR feature selection for the selected ML methods improve performance relative to a classifier built on all features or randomly sampled subsets. Permutation feature importance within the decorrelated MRMR features showed concordance in feature ranking between ML methods. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from DESeq2 analysis. Non-intersecting sets introduce opportunity to explore genes relevant to differentiating space flight exposed groups and implementing ML methods across existing NGS datasets may overcome sample size limitations.
Machine Learning↗
Transcriptomics-based Machine Learning (ML) Analysis Predicts Space-Exposed Murine Livers
Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), typically limits machine learning (ML) approaches in spaceflight studies that include radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNA-seq) data from six mouse liver GeneLab datasets (GLDS) (n ranging from 6 to 39 samples) from with a total of 81 spaceflight and ground-control samples to determine top features (i.e. genes) relevant to spaceflight including the effect of radiation exposure. RNASeq counts were normalized for each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. Redundancy and relevance were computed using the Pearson correlation and F-statistic, respectively. The top 100 MRMR features were used to predict spaceflight vs. ground-control samples using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) classifiers with 5-fold cross validation (CV). Principal component analysis (PCA) on the complete feature set versus the MRMR features shows separation between spaceflight samples and ground controls (Figure 1A). The ML-based gene sets were compared against differential gene expression results obtained with DESeq2 from individual GLDS. Using all features or randomly sampled subsets at matching set sizes with MRMR, a maximum classifier accuracy of 69% on the test set over 5 folds. For all classifiers, CV training using at least the top 30 MRMR genes show minimum 89% accuracy and 0.95 AUC value on the test set over 5 folds (Figure 1B). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 295 DEGs that overlap at least two studies and 13 DEGs that overlap three studies (Figure 1C). Set analysis between the top 100 MRMR features and the DEGs showed 47 genes that overlap at least one study and 24 genes that overlap two studies. Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism which may indicate these processes in the response to spaceflight stressors. MRMR feature selection for the selected ML methods improve performance relative to a classifier built on all features or randomly sampled subsets. Permutation feature importance within the decorrelated MRMR features showed concordance in feature ranking between ML methods. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from DESeq2 analysis. Non-intersecting sets introduce opportunity to explore genes relevant to differentiating space flight exposed groups and implementing ML methods across existing NGS datasets may overcome sample size limitations.
Machine Learning↗
Transcriptomics-based Machine Learning Analysis Predicts Space-Exposed Murine Livers
Limited sample sizes, high data dimensionality, and sensitivity to technical and biological variability of next generation sequencing (NGS), typically limits machine learning (ML) approaches in spaceflight studies that include radiation effects. However, pooling smaller studies while addressing intra- and inter-study variabilities allows for ML predictive modeling. Here, integration methods were applied to whole transcriptome shotgun sequencing (RNA-seq) data from six mouse liver GeneLab datasets (GLDS) (n ranging from 6 to 39 samples) from with a total of 81 spaceflight and ground-control samples to determine top features (i.e. genes) relevant to spaceflight including the effect of radiation exposure. RNASeq counts were normalized for each study, then merged and scaled across all datasets. Data dimensionality was reduced using a minimum redundancy maximum relevance (MRMR) methodology. Redundancy and relevance were computed using the Pearson correlation and F-statistic, respectively. The top 100 MRMR features were used to predict spaceflight vs. ground-control samples using Random Forest (RF), Support Vector Machine (SVM), and Linear Discriminant Analysis (LDA) classifiers with 5-fold cross validation (CV). Principal component analysis (PCA) on the complete feature set versus the MRMR features shows separation between spaceflight samples and ground controls (Figure 1A). The ML-based gene sets were compared against differential gene expression results obtained with DESeq2 from individual GLDS. Using all features or randomly sampled subsets at matching set sizes with MRMR, a maximum classifier accuracy of 69% was shown on the test set over 5 folds. For all classifiers, CV training using at least the top 30 MRMR genes show minimum 89% accuracy and 0.95 AUC value on the test set over 5 folds (Figure 1B). Baseline set analysis on differentially expressed genes (DEGs) identified using padj ≤ 0.05 show 295 DEGs that overlap at least two studies and 13 DEGs that overlap three studies (Figure 1C). Set analysis between the top 100 MRMR features and the DEGs showed 47 genes that overlap at least one study and 24 genes that overlap two studies. Over-representation analysis showed overlapping biological processes related to fatty acid and lipid metabolism which may indicate these processes in the response to spaceflight stressors. MRMR feature selection for the selected ML methods improve performance relative to a classifier built on all features or randomly sampled subsets. Permutation feature importance within the decorrelated MRMR features showed concordance in feature ranking between ML methods. A challenge of applying ML methods across heterogeneous NGS data is accounting for signal:noise. Here, signal validation across studies was shown by intersecting sets between top MRMR genes and DEGs from DESeq2 analysis. Non-intersecting sets introduce opportunity to explore genes relevant to differentiating space flight exposed groups and implementing ML methods across existing NGS datasets may overcome sample size limitations.
Machine Learning↗